Cascaded classifier for large-scale data applied to automatic segmentation of articular cartilage

Abstract

Many classification/segmentation tasks in medical imaging are particularly challenging for machine learning algorithms because of the huge amount of training data required to cover biological variability. Learning methods scaling badly in the number of training data points may not be applicable. This may exclude powerful classifiers with good generalization performance such as standard non-linear support vector machines (SVMs). Further, many medical imaging problems have highly imbalanced class populations, because the object to be segmented has only few pixels/voxels compared to the background. This article presents a two-stage classifier for large-scale medical imaging problems. In the first stage, a classifier that is easily trainable on large data sets is employed. The class imbalance is exploited and the classifier is adjusted to correctly detect background with a very high accuracy. Only the comparatively few data points not identified as background are passed to the second stage. Here a powerful classifier with high training time complexity can be employed for making the final decision whether a data point belongs to the object or not. We applied our method to the problem of automatically segmenting tibial articular cartilage from knee MRI scans. We show that by using nearest neighbor (kNN) in the first stage we can reduce the amount of data for training a non-linear SVM in the second stage. The cascaded system achieves better results than the state-of-the-art method relying on a single kNN classifier.

A review on segmentation of knee articular cartilage: from conventional methods towards deep learning

“…Another conventional technique is to perform a cascaded learning approach. Prasoon et al [172,173] proposed an extended cascaded classification framework originally introduced by [50]. This extended framework is known as a hierarchical classification scheme to segment femoral and tibial cartilage.…”

Femoral cartilage segmentation in Knee MRI scans using two stage voxel classification

“…The approach is compared to the state-of-the-art method that is based on one stage of nearest neighbor classification. We discuss the similarities and differences in segmenting femoral and tibial cartilages as well as the challenges faced due to the even higher amount of training data compared to [2]. Furthermore, we consider images of subjects scanned twice within one week and investigate the inter-scan reproducibility of the proposed classifier in comparison to a radiologist and the current state-of-the-art method.…”

“…An example of such a classifier is a non-linear support vector machine (SVM, [1]), where the training time scales worse than quadratically with the number of training data points. In our previous work [2], we presented a two-stage cascaded classifier approach to overcome this restriction. The proposed classifier was applied to segment tibial cartilage in low-field knee MRI scans and outperformed the state-of-the-art method.…”